For a while now I’ve had two jobs that don’t obviously belong together.

By day I design AI systems for organizations: agents, integrations, the unglamorous plumbing that turns a demo into something a company can rely on. The questions there are practical. Does it work? Is it safe? Will the people who have to use it trust it?

The rest of the time I’m an independent researcher working on the ethics of AI, reading Jacques Ellul and asking a less comfortable question: what does it do to us when we get very good at making systems that are powerful, efficient, and hard to say no to?

Most of what I read treats those as separate conversations. The engineers talk about capability. The ethicists talk about limits. I’ve come to think they’re the same conversation, and that the people who build these systems are exactly the ones who need to be having it.

So this is a place to think out loud. Some posts will be technical: patterns for agentic systems, lessons from real deployments, things I wish I’d known earlier. Some will be about the deeper question of what we’re actually doing when we build tools that act on our behalf. I’ll try to keep both in plain language.

If you’ve found your way here from LinkedIn, welcome. I’d rather have a conversation than an audience, so if something here provokes a thought, I’d like to hear it.